At the beginning of the design of any tool, molecular or otherwise, one needs to set the requirements that it needs to satisfy. We propose six principal characteristics (hallmarks) that a responsive probe should have for effective analyte detection:
fast response kinetics,
maximum signal difference,
initially a silent probe (off-on),
a decidedly binary response,
low toxicity (if considering in vivo applications)
A responsive probe must necessarily show an intrinsic reactivity to be susceptible to the target analyte. This requirement can become a serious hurdle in its design, in that the molecule may show residual spontaneous degradation in the absence of the analyte. This in turn likely results in the formation of the signaling molecule and thus in a false-positive signal, a well-recognized challenge in the design of responsive probes (for two examples, see Scheme 2 A,B).46 It is often difficult for the reader of scientific reports to assess the robustness of a given probe design from the furnished response data. If data are only given for the favorable case of high analyte activity or a large analyte-to-probe ratio, then the active transformation reaction may be much faster than the spontaneous degradation reaction so that no increase of a background signal needs to be feared during the monitoring period. In a more realistic setting of a heterogeneous sample with a spatially complex structure, the probe requires time to reach all sites (diffusion), and it may be in this very time period that it starts to generate a false-positive signal that seriously puts the active response result in doubt.
Scheme 2. Examples of nonmagnetogenic probes discussed in the light of hallmark satisfaction (gly=glucuronyl; r1=relaxivity).
Download figure to PowerPoint
A fast probe response is an obvious requirement. A responsive probe may show high, medium, or low reactivity toward the target analyte. Its signal may arise after one chemical step or after two or more. The consequences of slow response are dire: higher concentrations of the intact probe and/or the target analyte need to be established at the site where the analyte resides to observe the same signal intensity as for the case of a fast-responding probe at the same delay time. Simply increasing the delay time does not alleviate the situation: the diffusion rate of the activated probe remains roughly the same, and the signal from that particular site does not increase because of loss to other parts. A slow response does not exactly improve this situation when probes are considered that react through two steps or more: in structurally complex samples, the analyte residing in a particular place may chemically modify the probe as planned. However, the slow translation of this modification into a legible signal in the second step, if it is the rate-determining one (for an example, see Scheme 2 B),46a allows the intermediate to diffuse away from the site before “lighting up”. All these considerations can be summarized by the phenomenon of signal dilution that amounts to a decrease in detection sensitivity.
It is clear that no matter what the identity of the physical signal that serves to detect the analyte (the detection modality), a maximum emission intensity of the activated probe is always desirable. This ensures the highest difference between the “before” and “after” state of the responsive probe (for an opposite example, see Scheme 1 C).47 In the particular case of a magnetogenic probe this would imply the generation of a high number of unpaired electrons at the nucleus.
It is highly desirable to design a probe that is initially silent (Scheme 2 A,B).48 The absence of any signal in the absence of analyte simplifies the interpretation of the image or detection result enormously. It also makes the dream of achieving analyte quantitation much more realistic. It has been repeatedly stressed in the case of responsive probes for MRI that already emit a sizeable signal before they encounter the target that unambiguous image interpretation would require knowing the concentration of converted and unconverted probe at the site of interest.8a, 49 Without this knowledge, it cannot be ruled out that the observed signal is simply the result of an accumulation of the untouched probe for physicochemical reasons (biodistribution, diffusion, lipophilicity, charge, etc) without the presence of any analyte at all. For the same reasons outlined above, the opposite scenario of an onoff probe31, 50 is less attractive (for an example showing this tendency,47 even though the process does not reach an off state, see Scheme 2 C).
Depending on the nature of the activation reaction (reversible versus irreversible reaction), the choice for one side or the other of the associated chemical equilibrium may be more or less decided. In other words, it should be avoided that only part of the population of intact probes is converted by the analyte, or that only part of the population of activated probes emits a signal (for comparison see Scheme 1 A,B). The majority of detection schemes covered by Table 1 enter into this category. In contrast: a switch of the entire population from 0 to 100 % should be the goal (true for all examples in Scheme 2). Otherwise, the same complications arise that were already mentioned for the previous hallmark of probe design.
3.2. Choice of Magnetogenic Core
Magnetization is the density of magnetic dipole moments induced by the presence of an external magnetic field generated by a permanent magnet surrounding the sample. Such magnetic moments are the result of the spin of electrons and nuclei as well as electron orbital movements. One speaks of an electronic and a nuclear angular momentum, and the former is composed of a spin angular momentum and an orbital angular momentum. The nuclear angular momentum can be neglected because it is approximately 1000 times smaller than the electronic one. A magnetic moment can be detected when unpaired electrons are present in the atom. Compounds that possess such unpaired electrons are called paramagnetic and are attracted by the external magnetic field, while those that are diamagnetic are repelled. The overall spin S generated by these electrons is the major contribution to the magnetic moment for elements of lower atomic numbers Z. The contribution by the electron orbital movement becomes increasingly important for elements of higher atomic numbers. Finally, the electronic spin and the orbital spin can interact (spin-orbit coupling), and this also contributes to the magnetic moment; its level depends on the element and the external magnetic field. Spin-orbit coupling can be quenched to different degrees by changing the electronic configuration and symmetry through the choice of the surrounding molecular scaffold. While it is very small for first-row transition-metal complexes, it contributes more significantly in heavier d-block metal and lanthanide complexes, and can exceed the spin-only contribution in actinide complexes.
A probe must possess a section that is capable of emitting a detectable signal. For a magnetically responsive probe this unit cannot simply possess unpaired electrons, and thus a spin and a magnetic moment, but rather it has to be capable of adopting two different magnetic states depending on its interaction with the analyte (or stimulus). Three types of structurally analogous pairs may be envisaged where this is possible: a) a duo where one compound is an organic radical, b) an internal redox duo experiencing electron transfer, and c) a duo that consists of two spin states (generally a low-spin and a high-spin state).
a) Radicals: Although paramagnetic organic radicals may be generated from a diamagnetic precursor and thus fulfill this requirement, they are usually rather unstable. A few exceptions are fairly stable in solution (e.g. spin labels) and have occasionally been considered for the design of responsive organic compounds that become paramagnetic radicals or lose this quality.51 Examples of organic radicals that change their level of paramagnetism are also common in solid-state paramagnetic/diamagnetic switches that respond to physical stimuli.10, 52 Upon heating or irradiation, the paramagnetism of the radical becomes quenched because of a change in its relative position in the crystal lattice. These switches are based on a pronounced cooperative effect and their mode of action can thus not be transferred to the solution phase unless some sort of self-assembly can be achieved.53 On the other hand, the quenching of two radicals contained in one isolated molecule in solution has been reported on a number of occasions, but only as a result of physical stimuli (summarized in Ref. 10) or a pH change,51a but not for other chemical stimuli. Some examples were reported where a photoswitchable moiety caused a change in the intramolecular communication between two radicals, thereby leading to magneto-modulation.54
b) Intramolecular redox reactions: A redox-active analyte may change the magnetic properties of a coordination compound by exchanging electrons directly with the magnetogenic core in an intermolecular redox reaction. Such a process makes it rather difficult to confer any detection specificity for a particular analyte onto the probe, but should rather serve to characterize a general redox potential, which is of great interest for biological research. Any intermolecular electron flow may also cause undesired side reactions. Another way of modifying the magnetic quality of a molecular entity would be an intramolecular electron-transfer process, namely, a redox reaction caused by an external stimulus that is not redox-active. Most transition-metal ions that can adopt at least two stable oxidation states are suitable candidates, and even radicals may be considered. An intramolecular redox-driven change in magnetism has been widely reported for both the solution and the solid state, but almost exclusively as a result of a physical stimulus such as light, temperature, or pressure.10, 13, 14a Numerous examples from two types of electron transfer have been reported: 1) ligand-to-metal transfer, also referred to as valence tautomerism (VT), including the classic semiquinone-catechol cobalt complexes, where a CoIII LS center becomes a CoII HS one (Δeunpaired=3); and 2) metal-to-metal transfer in polynuclear complexes, where Fe-Co, Fe-Fe, Fe-Ni, and other couples are bridged by ligands such as cyanide, including the classic example of prussian blue [Fe4[Fe(CN)6]3]. Molecules that operate in this fashion must ensure the possibility of electron transfer between the two redox-active portions of the molecule, that is, between the highest occupied molecular orbital (HOMO) of the donor portion and the lowest unoccupied molecular orbital (LUMO) of the receptor unit. The opportunity thus arises of making compounds that respond to the presence of a chemical analyte should they interact in such a fashion so as to invert the HOMO–LUMO relationship, thereby leading to electron transfer and a change in the oxidation state of the metal center. If it is coupled with a change of the spin state of a central metal ion, then the change in magnetization can be very high. Only a few examples exist where the intramolecular electron transfer is caused by a purely non-redox-active stimulus.34, 51a, 55
c) Low-spin/high-spin switching: Significant changes in magnetization can be obtained with a probe that switches from a low-spin (LS) to a high-spin (HS) state as a response to ligand modification. Ligand field theory stipulates that initially degenerate d orbitals experience a splitting into different energy levels as a result of the approach of the ligand(s) (Figure 1 A). Some d orbitals are more affected than others because the approach of the ligand is directional, that is, metal and ligand orbitals of the same symmetry interact more strongly. By modulating the field of these ligands, one may induce the complex to adopt either a LS or HS state. Field splitting depends as much on the σ-donating as on the π-accepting capacity of the ligand(s). The presence of a low-lying antibonding π orbital (π*) is necessary for a given ligand to be a good π acceptor; the resulting increase in the ligand field is referred to as π back-bonding. As we shall see in Section 3.3 (Table 2), imine-type ligands (containing sp2-configured nitrogen atoms) cause a particularly strong splitting of the ligand field because of their high σ-donor and π-acceptor qualities.
Figure 1. Field splitting (ΔE, length of double arrows) caused by ligands of different strength and resulting d electron configurations for low-spin and high-spin complexes of first-row transition metals in terms of the presence or absence of an off-on relationship, and the magnitude of the signal difference (eHS-LS); dashed arrow: 1.23 of ΔE(octahedral); gray dashed arrow: 4/9 of ΔE(octahedral).
Download figure to PowerPoint
Table 2. Hexadentate ligands giving rise to binary FeII complexes.
|Podand||Base system||N hybridization||Spin state|
|branched||Py3tame (C-branched)||6 N-sp2||LS59|
| ||tptMetame (C-branched)||3 N-sp2, 3 N-sp3||SCO60|
| ||Py3tren (N-branched)||6 N-sp2||LS61|
| ||trimethylenediamine||4 N-sp2, 2 N-sp3||LS62|
| ||2,5,8-triazanon-1-ene||4 N-sp2, 2 N-sp3||LS63|
| ||cis,cis-1,3,5-cyclohexane||3 N-sp2, 3 N-sp3||LS64, 41|
|macrocyclic||triazacyclononane||3 N-sp2, 3 N-sp3||LS65a,b|
|bicyclic||bicyclo[7.5.5]nonadecane||3 N-sp2, 3 N-sp3||LS66a,b|
| ||bispidine||4 N-sp2, 2 N-sp3||LS67a,b|
The nature of the metal ion also plays a role in field splitting: the higher the oxidation number, the higher the energy splitting, which favors the LS state. Only octahedral (or pseudo-octahedral) complexes with a d4 to d7 configuration can effectively adopt either a LS or a HS state (Figure 1 B–E), with the exception of a d8 configuration (Figure 1 F), should the corresponding complex change its coordination chemistry from square planar (LS and diamagnetic) to tetrahedral or octahedral (both HS and paramagnetic) in the process.11a, 22 A tetrahedral geometry normally favors the HS state for the first-row transition metals because the ligand-induced splitting energy is only 4/9 of that observed for an octahedral coordination geometry, and thus too small to overcome the spin-pairing energy. On the other hand, second- and third-row transition metals are mostly found in the LS state because of strong field splitting. Lanthanides are not suitable either, as they are always found in only one spin state. Their paramagnetism is also as much dependent on spin-orbit coupling as on their number of unpaired electrons. No matter if the strategy of internal electron transfer (b) or that of LS-HS switching (c) is pursued, the choice of the right metal is of pivotal importance.
The d4 to d8 metal ions have advantages and drawbacks when viewed from the hallmarks of probe design listed above. In fact, the electronic configuration not only has a critical influence on the maximal signal difference attainable (eHS-LS, difference in unpaired electrons before and after activation), it also determines whether a true off-on activation mode is feasible. In the best case, the LS state should thus show no unpaired electrons at all (diamagnetic, off, spin=0), which is only possible for an octahedral d6 and a square-planar d8 configuration. On the other hand, a maximal signal difference is only ensured for d5 and d6 configurations (eHS-LS=4 e); others give only half this difference (2 e). In theory, the d6 ions FeII and CoIII can be considered optimal, with the d5 ions FeIII and MnII, as well as d8 square-planar NiII as possible alternatives. In practice, however, MnII suffers from such a high pairing energy that it is very difficult to attain its LS state,56 and CoIII is found in the LS state in the vast majority of cases; ligands producing a particularly weak ligand field (fluoride) are required to turn it into its HS state (see CoF63−). Thus, no practical ligand system can be found that may cause CoIII to adopt both of the spin states as a result of slight structural modifications.
Another hallmark of probe design is robustness, and the remaining contenders (FeIII, FeII, and NiII) show various characteristics in this regard. Ferric complexes (FeIII) of both spin states can be quite stable. However, although a good number of ligand systems is known that leads to either LS or HS complexes, this selection is more limited than that known for ferrous complexes (FeII).11a This results in the difficulty of identifying a duo of structurally related ligands where the corresponding ferric complex adopts a decidedly LS state while the other leads to a fully HS one. The hope of observing the maximum signal difference (hallmark) promised theoretically by a ferric system is thus diminished. On the other hand, ferrous complexes in the HS state may or may not show a tendency towards oxidation, and the ligand system has thus to be designed so as to minimize it. Importantly, a large range of different ligands are at our disposal that cause FeII to adopt one of the two spin states.11a, 57 We thus have within our grasp a sharp magnetogenic response caused by ligand modification under the influence of a chemical analyte. Furthermore, LS ferrous complexes are highly stable and show a certain level of kinetic inertness.58
Finally, NiII complexes, although having a true “off status” for the initial probe and suffering from a mediocre signal difference (eLS-HS=2), adopt a square-planar geometry that is not only unstable in water but also in organic solvents that contain a complex mixture of nucleophiles.44a,c So far, no convincing design has been reported where the LS/off state of NiII chelates was explored for probing/analysis in aqueous or complex samples. In addition, only a limited range of ligand systems are available to modulate the spin state of NiII complexes. In the following we, therefore, direct our attention towards FeII systems.
3.3. FeII LS/HS Duos
Well-established methods for tuning the magnetic state of iron(II) in its complexes comprise:
variation of the nature of the coordinating atom (mainly, N, O, S),
variation of their nucleophilicity/basicity/σ-donating capacity by, for example, decoration with electron-withdrawing or -donating groups,
variation of the hybridization of the N atom (imine versus amine=aromatic versus aliphatic=π back-bonding or not),
impeding optimal orbital overlap by introducing steric clash,
presence or absence of the macrocyclic effect, and
switching between five- and six-membered chelate rings.
In applying these criteria, we will not overlook the primordial requirement of solution-phase stability. Indeed, FeII complexes with hexadentate ligands “are known to have stability constants of the order of 1025”,20b and “most FeII SCO systems based on multidentate ligands are so stable that ligand dissociation does not interfere with the spin equilibrium even in polar solvents”.20d However, “ligand dissociation and replacement reactions are more likely to occur for complexes of mono- and bidentate ligands, but even for multidentate ligands replacement of a single chelate arm has been observed.”20c These thermodynamic considerations should not hide the fact that for solution-phase applications in complex samples, especially biological ones, kinetic factors will almost always override them:68 in fact it is the exchange equilibria with other, abundant metal ions and the rates at which they are established that will decide whether slow or rapid probe degradation occurs.69 What is thus expected to work in aqueous media is a ligand structure for which an N6 and an N5 ligand can be constructed, because even changing the coordination motif from N6 to N5O1 during probe response (Scheme 3) will make a switch to a HS system highly likely, provided the oxygen atom is neither sp2-configured (part of a carbonyl group) nor part of triplet oxygen (see BOLD fMRI). From this it can be concluded that if an octahedral ferrous chelate can be identified that is fully low spin at room temperature and in aqueous solution, it is almost guaranteed that a high-spin version of it can be obtained under the same conditions if one arm becomes decoordinated or cleaved off.
These considerations prompted us to examine the literature for cases of binary ferrous complexes that are fully LS in solution, preferably in water (Table 2). By studying the broad review by Halcrow on ferrous complexes with multidentate N ligands one may identify several LS-HS duos; for some of them both derivatives were already characterized magnetically, for other duos only one has so far been characterized.57 However, that review also illustrates that rather few hexa-/pentadentate systems do in fact ensure a robust low-spin state and so might be attractive for the present task. Multidentate ligands that give rise to binary low-spin ferrous complexes comprise 1) branched podands, 2) macrocyclic podands, and 3) multicyclic rigid podands (Table 2). Branched podand-based N6-FeII chelates can be subdivided into a) tripodal ligands branched by a single atom, b) alkyl-diamine-based podands, and c) two unique cases with special branching units. Examination of examples entering into category 1 a shows that a simple branched N6 ligand (tptMetame) cannot force the iron center to adopt the LS state if “only” three nitrogen atoms are sp2-configured.60 Only ligands with six imine-type nitrogen atoms have, currently, yielded LS complexes: While Py3tame59 suffers strain (its LS state can thus be considered weak), Py3tren61 exhibits a tendency for decoordination of the pendent arm,70 and so despite the presence of six coordination sites of high ligand field, their suitability for the design of a magnetogenic unit remains limited.
Alkyl-1,2-diamines (category 1 b) lead to intermediate paramagnetism in FeII complexes despite them generally being equipped with four sp2-configured nitrogen atoms;71 they have also been observed to show decoordination of the pendent arm.62 Expansion of the ethylene bridge by one further methylene group reliably leads to low-spin complexes (see Scheme 3, 162 and its HS analogue 2).72 Three ligands were found to belong to category 1 c (unique branched ligands). Only two of them (2,5,8-triazanonene and r-1,c-3,c-5-triaminocyclohexane (tach); 7) caused the corresponding FeII complexes to adopt a LS state, but have the advantage of doing so not only in the solid state but also in solution, including water.41, 63, 64 They exhibit four and three sp2-configured nitrogen atoms, respectively, out of six. For category 2 (macrocyclic podands), it should be mentioned that many phthalocyanine FeII complexes were obtained in the LS state, but as the ligands are tetradentate the octahedral complexes cannot be binary. The rigid planar nature of these N4 macrocyclic ligands precludes the design of N6 derivatives that lead to hexacoordinate, binary FeII complexes. It has been shown that the required chelate ring sizes are simply too large to form. Accordingly, it is not apparent how these types of ligands may be useful in the design of robust magnetogenic probes for use in the solution phase, apart maybe from their synthetically challenging incorporation into a multicyclic system. On the other hand, the tripyridylmethyl derivative of the N3 macrocycle triazanonane (tacn) has led to the preparation of a fully LS FeII complex (three imines/three amines, Scheme 3, 3) in the solid state.65 This LS state is maintained in various aqueous media,73 while the corresponding complex lacking one pyridylmethyl arm74 is of course HS (4).73 The HS and LS versions were shown to be visible and invisible, respectively, in MR images of a live mouse.75 As we shall see in Section 4.3, this system can be effectively transformed into two types of magnetogenic probes that respond to various chemical stimuli. A reported bicyclic N5 ligand can be assigned to category C. Its dark-red ternary ferrous complex (5) was recrystallized from acetonitrile and proved to be LS in the solid state.66 The sixth coordination site was occupied by an acetonitrile ligand. No hexadentate ligand or its corresponding complex has yet been reported, but there is no reason why this system may not also serve as a promising target for the design of a robust magnetogenic probe by the introduction of another pendant arm (5). We recently reported new N6 members for category C, namely two bicyclic, rigid, and hexadentate ligands of the unique class of the bispidines (Scheme 3).67b Multidentate bispidines have previously been reported to form highly stable complexes.76 We prepared new bispidine ligands that led to the discovery of the first two binary LS FeII chelates (9) for this large class of bicyclic structures. Their LS nature and high stability was confirmed in water and organic solvents at ambient temperature, and their magnetism was explored exhaustively.67b Simple removal of one coordinating arm leads to a fully HS system in water (10; 5.0 μB).67 A new robust off-on duo of ferrous complexes in aqueous media at room temperature has thus become available. While macrocyclic as well as bicyclic ligands (categories 2 and 3) may generally require significantly higher synthetic efforts, it is important to note that the reported hexadentate bispidine ligands can be efficiently prepared on a 10 gram scale.67b
In conclusion, this literature survey reveals that binary LS FeII complexes are not that numerous after all. It also teaches us that simple branched N6 ligands do not appear to impose a LS state if they do not exhibit more than three imine-type coordination sites. For this reason, and for reasons of solution stability, macrocyclic or bicyclic N6 ligands should be preferred. Some of the promising LS-HS duos for the design of a putative magnetogenic probe are presented in Scheme 3.